Method: We used Surveillance, Epidemiology, and End Results (SEER) cancer registry data linked with Medicare claims for patients aged 65+ diagnosed with stage IV PCa between 2000 and 2007. The physician visit index (PVI), which is derived from the Herfindahl index, was used to measure multi-specialist contact across physician types including urologist, radiation oncologist, medical oncologist, nuclear medicine physician, or primary care physician within 6 months (6m) following diagnosis. The PVI ranged from 0 (multiple types of physicians seen) to 1 (single type of physician seen). We estimated a two-level (i.e., patient- and physician-level) hierarchical linear regression model of the 6m PVI adjusted for patient, physician, and geographical factors. The intraclass correlation coefficient (ICC) measured the proportion of variance in the PVI explained at the physician level.
Result: The sample included 2,885 men diagnosed with stage IV PCa who saw 656 unique physicians across 15 SEER registries. The average age was 76 years and the median (interquartile range) for the 6m PVI was 0.54 (0.26). Based on the ICC, 7.5% of the variation in 6m PVI was explained at the physician level (p<0.0001). Statistically significant patient-level factors associated with a lower PVI (more types of physicians seen) included urban residence, pre-diagnosis use of preventive services, 6m treatment receipt, and comorbidity. Physician-level fixed effects such as age, years in practice, and specialty were not statistically significant however, practicing in a non patient care setting was associated with a lower PVI (coefficient= -0.05; p=0.04) compared to practicing in a group setting. The 6m PVI was higher (fewer types of physicians seen) among patients in the West region (coefficient= 0.01; p=0.02) compared to patients in the Northeast region.
Conclusion: The types of physicians seen by patients following diagnosis of advanced PCa is influenced by patient, geographical and physician-level factors. The physician characteristics explored in this study did not fully explain the variation in PVI attributed to the physician, suggesting that additional investigation of physician-level factors is needed.